Configuration

The chart README lists every value. These are the ones most people change.

Helm values

Value	Default	What it does
install.pin	""	Show one install only: its namespace/name or install ID. Empty shows all.
httpRoute.enabled	false	Create an HTTPRoute to the UI. Read the trust model first.
httpRoute.parentRefs	[]	The Gateway(s) the route attaches to.
httpRoute.hostnames	[]	Hostnames for the route.
diagnostics.events	true	Watch Warning Events on controller and proxy Pods.
diagnostics.crds	true	Read Gateway API CRD metadata, for version-skew checks.
diagnostics.proxyStats	true	Read Envoy stats from each proxy Pod on port 9091.
diagnostics.endpoints	true	Watch Services and EndpointSlices, for backends without ready endpoints.
diagnostics.controllerLogs	false	Read controller logs for derived signals. Adds a pods/log Role in each namespace of diagnostics.controllerLogsNamespaces.
diagnostics.controllerLogsNamespaces	[kgateway-system]	The namespaces that get that Role. List every namespace that runs a kgateway controller.
delegation.edgeCap	0	Delegation edges computed per install; 0 keeps the default of 10,000.
installs.classOnly	false	Also show a GatewayClass whose controller Snellius does not recognise, from Gateway API status alone. From v0.6.0.
extraArgs	[]	More flags for the binary, for example {--poll-interval=10m}.
resources	50m CPU, 128Mi requested; 512Mi limit	Container resources.

Turning events, crds or endpoints off also drops its RBAC rule; proxyStats needs no RBAC, only network access to port 9091. Checks that need a disabled input read “not available”, and the install’s state does not change.

Binary flags

Set these through extraArgs, or pass them when you run the binary yourself.

Flag	Default	What it does
--poll-interval	30s	The longest time between recomputes. Snellius recomputes on every change it watches, debounced to about once a second; this is only the safety timer.
--metrics-interval	10s	How often the controller’s metrics are scraped.
--delegation-edge-cap	10000	Delegation edges computed per install. A higher cap needs more memory.
--install		Same as install.pin.
--listen	:8080	Listen address. Outside a cluster the default is 127.0.0.1:8080.
--kubeconfig		Kubeconfig path. Empty uses the in-cluster config.

Run it outside the cluster

To try Snellius against your current kubeconfig context, build it from source:

git clone https://github.com/DuncanDoyle/snellius && cd snellius
make build
./bin/snellius --kubeconfig ~/.kube/config

It serves on http://127.0.0.1:8080, with your kubeconfig’s permissions.